EMxDesign: A Genetic Algorithm for High Affinity Drug Design.

As pathogens are rapidly evolving to evade modern drugs, it is vital to design efficient, optimal treatments in a timely manner. One way to achieve this is by mimicking the process of evolution when designing drugs most fit to combat pathogens of interest through the use of genetic evolutionary algorithms. In this work, we introduce EMxDesign (Element Molecule Exchange Drug Design), a genetic algorithm that aims to optimize drug-pathogen affinity by inducing single element mutations in cross-overs of existing molecular drug structures. Using this approach, EMxDesign generates high affinity drug molecules while circumventing problems of existing genetic algorithms for drug design such as requiring specific hyperparameter tuning for different protein targets. We used EMxDesign to generate drug molecules for eighteen SARS-CoV-2 protein pocket targets, and found after just one iteration our newly generated molecules had significantly higher mean affinity for thirteen of our target pockets. We are currently refining this framework to launch an application to aid in drug development at scale.